Bowen Yu

HE TURNS RAW MODELS INTO INSTRUCTION-FOLLOWING ASSISTANTS

Yu earned his PhD from the Chinese Academy of Sciences, joined Alibaba as a top young researcher, and led Qwen's post-training — the pipeline stage that turns raw pretrained models into instruction-following assistants. He was second-listed author on the Qwen2.5-1M report that stretched post-training to a million tokens of context.

1. 2017

   HE LEAVES UNIVERSITY AS A TOP GRADUATE

   Yu's biography lists him as an Outstanding Graduate of Hunan Province in 2017 before his doctoral work in information engineering.

   SOURCES: Bowen Yufirst-party

2. 2019

   THE ALIGNMENT RESEARCHER STARTS WINNING NATIONAL AWARDS

   During doctoral study, Yu received a national doctoral scholarship and a top student honor from the University of Chinese Academy of Sciences.

   SOURCES: Bowen Yufirst-party

3. 2022

   PHD FINISHED. ALIBABA CALLS HIM A STAR.

   Yu completed his PhD at the Institute of Information Engineering, Chinese Academy of Sciences, and received Alibaba's top-one-percent Alibaba Star award.

   SOURCES: Bowen Yufirst-party

4. Sep 28, 2023

   HE HELPS TURN QWEN INTO AN INSTRUCTION-FOLLOWING MODEL

   Yu co-authored the first Qwen technical report, including its aligned Qwen-Chat models, and later became the team's post-training lead.

   SOURCES: Qwenreported · Bowen Yufirst-party

5. Jul 15, 2024

   POST-TRAINING SCALES WITH QWEN2

   Yu was among the core authors of Qwen2 as the team expanded instruction-tuned models across a much broader size range.

   SOURCES: Qwenreported

6. Jan 26, 2025

   QWEN'S POST-TRAINING STRETCHES TO A MILLION TOKENS

   Yu was the second listed author on the Qwen2.5-1M report describing long-context pre-training and multi-stage supervised fine-tuning.

   SOURCES: Qwenreported